1
Can Sterilized FX Purchases under Inflation Targeting Be Expansionary?
Márcio G. P. Garcia12
Visiting Scholar, Sloan School – MIT and NBER
Associate Professor, Department of Economics - PUC-Rio
This version: September 3, 2013
Abstract:
Contrary to conventional wisdom, sterilized foreign exchange (FX) purchases under monetary
regimes with an interest rate rule, e.g., inflation targeting, generally increase aggregate demand.
We develop a simple model with a credit channel to argue that sterilized FX purchases, by
funding bank credit, end up increasing aggregate and money demand, while expanding loans
and deposits, and reducing the loan interest rate. Therefore, restoring the interest rate to the
level prior to the sterilized FX purchase is not sufficient to prevent an expansionary effect; the
new money market equilibrium, at the same interest rate, will entail a larger money supply,
higher output and greater money demand. Recent Brazilian evidence is reviewed, showing that
this effect may be empirically relevant. If this is the case, inflation targeters may have another
reason to be concerned when conducting FX sterilized interventions, besides their high cost and
controversial effectiveness in preventing nominal appreciation. Even if FX sterilized purchases
are effective in preventing nominal appreciation, they generally boost activity and inflation,
thereby appreciating the real exchange rate.
JEL Codes: F3, F4, E5
Keywords: Sterilized Interventions, Capital Flows, and Inflation Targeting
1 E-mail: [email protected] . Website: www.econ.puc-rio.br/mgarcia. 2 I thank Rogério Werneck, Eduardo Loyo, Linda Goldberg, Marcelo Abreu, Pablo Kurlat, Eduardo Levy-Yeyati, Samuel Pessoa, Alexandre Schwartsman, Mario Mesquita, Armínio Fraga, Afonso Bevilaqua, Luis Catão, Ricardo Velloso, Marcos Chamon, Irineu de Carvalho, Otaviano Canuto, Lucio Sarno, Gustavo
Loyola, Fabio Giambiagi, Jonathan Ostry, Diogo Guillen, Pedro Maia, Bernard Appy and Nicholas Hope for invaluable comments and suggestions, as well as Alessandro Rivello, Carolina Machado, Bruno Balassiano and Guido Maia for expert research assistance. All errors are mine.
2
Index
1. Introduction .....................................................................................................................3
2. A Simple IS-LM Model with Two Assets ..............................................................................6
3. Effects of Sterilized Interventions ......................................................................................9
4. Different Types of Capital Inflows and Policy Measures ..................................................... 15
5. FX Sterilized Interventions in Brazil and Money ................................................................ 17
6. Concluding Remarks ....................................................................................................... 26
7. References ..................................................................................................................... 29
3
1. Introduction
Until recently, international financial markets were very liquid. The abundant liquidity,
together with the good prospects of many emerging markets, drove massive capital flows to
these economies.
Several emerging markets, like Brazil, have been conducting monetary policy in an inflation
targeting framework. Although this framework prescribes a free floating exchange rate, the
exchange rate appreciation caused by capital inflows is increasingly being seen as extremely
detrimental to long term growth. The specter of Dutch disease has often been invoked.
Therefore, several forms of intervention in exchange rate markets without violating the open
economy inconsistent trinity have been attempted. The main ones are controls on capital
inflows and foreign exchange (FX) sterilized purchases.
In a country with extremely high interest rates, as it was (and still is, although not so high)
the case of Brazil, the foreign exchange (FX) reserves purchased through sterilized
interventions are very expensive, thereby generating high fiscal costs.3 The theoretical and
econometric evidence as to the effectiveness of sterilized purchases of FX in depreciating the
home currency is also very mixed.4 Nevertheless, sterilized interventions have been conducted
in Brazil for a considerable period of time, generating a very large volume of foreign reserves
(around USD 370 bi, or over 16% of GDP).
Despite its flaws, sterilized purchases of FX are widely believed to have no effects on
economic activity. To illustrate this point, imagine an open economy with unemployment at
NAIRU, GDP growth at the normal rate, the real interest rate at the neutral rate and the
inflation rate equal to the inflation rate target.
Suddenly, capital starts to flow into this economy because oil, for example, has been found
or because risk aversion has decreased worldwide. The inflation-targeting-monetary-policy
maker decides to fully sterilize the capital inflow. Under an inflation targeting regime, this
means purchasing all the FX inflow with domestic currency, thereby lowering the nominal
interest rate, while simultaneously conducting contractionary open market operations that
restore the previous nominal interest rate.
Are such sterilized interventions under inflation targeting expansionary?5 Most
economists, at least those I have interviewed, will answer in the negative. This paper argues
that the answer is most likely to be positive.
3 According to Credit Suisse, “... evaluating the sterilization cost, according to the main market indicators for the cost of rolling over debt, the fiscal cost of carrying the reserves would be approximately 1.4% of GDP per year from 2004 to 2010. In the 12-month period through June 2011, the cost of carrying the reserves would be 2.7% of GDP, nearly equivalent to the central government´s primary surplus in the same period” (Credit Suisse, 2011). 4 Dominguez and Frankel (1993), Sarno and Taylor (2001) and Neely (2005) provide classical surveys. Engel (2013) provide a recent evaluation of the econometric results, ending with the following caveat:
“despite many empirical studies, it is not clear yet whether sterilized intervention meets the same criteria that regulators use to decide whether to approve a cancer drug—that is safe and effective” (p. 39). 5 The question abstracts from possible effects on the exchange rate.
4
Policy-makers in developing countries have complained about capital inflows’
expansionary effect on credit aggregates. The Brazilian central bank, for example, considered
that: “… the fragility in some mature economies, combined with favorable perspectives for the
Brazilian economy, has determined an inflow of foreign resources, part of which has been going
to the credit market. In this sense, the excess of external inflows may weak (sic) the credit
channel, smooth its contribution to the aggregate demand moderation, as well as cause
distortions in the price of domestic assets” (Central Bank of Brazil, 2011). The Chilean central
bank warned: “… the main risks for financial stability associated with larger gross capital
inflows include the generation of currency and maturity mismatches, credit booms that lead to
a deterioration in loan quality, and local asset price misalignment” (Central Bank of Chile,
2011). The Turkish central bank admonished: “… in emerging economies, short-term capital
flows and rapid credit growth feed macro financial risks. … . The major risk factor for emerging
economies is the macroeconomic imbalances driven by rapid capital inflows. Central banks of
emerging economies continued to implement macroprudential measures to contain the
potential adverse effects of capital flows” (Central Bank of Turkey, 2011).
All these central banks have adopted the inflation targeting regime. They also intervene in
exchange rate markets through sterilized interventions. Therefore, if they are complaining
about the expansionary effects of capital inflows on credit aggregates, sterilized interventions
are not being effective in isolating the real economy from capital flows. However, current
models6 have no such expansionary effect.
In this paper, it will be shown that sterilized FX purchases, even if they are ineffective in
depreciating the exchange rate, do not immunize the domestic economy from the
expansionary effects of capital inflows, thereby justifying the policy-makers contentions.
Policy-makers, however, may be displeased to learn that, in order to counteract the
expansionary effect of sterilized FX purchases, contractionary policies (fiscal and/or monetary)
must be conducted. The idea that, by lowering interest rates, less capital will flow into the
country, thereby mitigating the expansionary effects of capital inflows is false. This is because
the capital inflows attracted to profit from high domestic government bond yields are not the
ones that generate the expansionary effect. The expansionary effect is generated by capital
that enters the country to finance aggregate demand expansion.
6 Signaling models could provide a rationale for sterilized FX purchases being expansionary. According to the signaling mechanism, those sterilized interventions would be a way for the Central Bank to signal future reductions in interest rates. As Obstfeld and Rogoff (1996) recognize, “… there certainly seem to have been episodes in which sterilized interventions, when concerted among large groups of countries, have clarified governments’ views on exchange rates and shifted market opinion…” about macroeconomic policies. However, the inflation-targeting (IT) framework has many channels through which the central bank may communicate its intentions to markets: monetary policy committee (MPC) minutes, inflation reports, etc. In fact, increased transparency and accountability are believed to be key improvements of IT over previous monetary policy regimes (Mishkin, 2000). It is very unlikely that any central bank that adopts IT would resort to sterilized interventions to signal a change in monetary policy.
Furthermore, it will be shown that, in Brazil, after the increase in sterilized purchases, the basic interest rate was raised, not lowered, as well as the other contractionary monetary quantitative measures (e.g., increases in reserve requirements) taken.
5
The next section reviews a simple model à la IS-LM with a banking sector that introduces a
new asset, credit.7 In Section 3, the model is extended to account for sterilized interventions.
The model shows that sterilized interventions under inflation targeting will, in general, be
expansionary. This result follows from two key features of the model: the existence of two
interest rates, the bond interest rate and the loan rate, as well as a portfolio effect, which
makes the banks increase loans, and reduce the loan rate, when their liabilities grow due to
the capital inflows. Section 4 develops elaborates on the model to conclude that FX inflows,
and therefore capital inflows, are not homogeneous as to their effects in the credit market.
This distinction is key to appreciating why lowering interest rates in times of high capital
inflows, with the aim of deterring these inflows, as Turkey did in 2010, may be ineffective and
fuel the credit market even more. Section 5 presents empirical evidence from Brazil supporting
the view that sterilized interventions under inflation targeting are expansionary. Finally,
section 6 concludes with a discussion of the policy implications of the expansionary effects of
sterilized interventions under inflation targeting.
7 Bernanke and Blinder (1988).
6
2. A Simple IS-LM Model with Two Assets
To illustrate how sterilized FX purchases under inflation targeting may be expansionary, we
resort to a simple IS-LM model with two assets, akin to the one developed by Bernanke and
Blinder (1988), henceforth referred to as the BB model.
In models inspired by the traditional IS-LM model, “… loans and other forms of customer-
market credit are viewed as perfect substitutes for auction-market credit (“bonds”)” 8. In the BB
model, a third asset, loans, is added to money and bonds.
Borrowers and lenders observe the relevant interest rates (i on bonds, and on loans) and
decide how to allocate their wealth. The demand for loans is, therefore, represented by
equation (1), where y (GNP) “… captures the transaction's demand for credit”9:
𝐿𝑑 = 𝐿(𝜌, 𝑖, 𝑦) (1)
Loans supply is performed through the banking sector. To understand how it works,
Figure 1 displays the simplified balance sheet of the representative bank, which is analogous to
balance sheet of the entire banking sector.
Figure 1: The Simplified Balance Sheet of a Representative Bank
Bank Balance Sheet Assets Liabilities
R (bank reserves) D (deposits)
Bb (bonds)
Ls (loan supply)
Bank’s assets are bank reserves (R), bonds (Bb), and loans (Ls). Bank’s liabilities are deposits
(D). Bank reserves (R) are composed of required reserves (τ.D) plus excess reserves (E).
Therefore, from the bank’s balance sheet:
𝐵𝑏 + 𝐿𝑠 + 𝐸 = 𝐷(1 − 𝜏) (2)
The portfolio shares of bonds (β), loans (λ) and excess reserves (ε), β+λ+ε=1, are
determined according to returns (zero for excess reserves):
𝐿𝑠 = 𝜆(𝜌, 𝑖)𝐷(1 − 𝜏) (3)
𝐵𝑏 = 𝛽(𝜌, 𝑖)𝐷(1 − 𝜏) (4)
8 Bernanke and Blinder (1988), p. 435. 9 Bernanke and Blinder (1988), p. 435.
7
𝐸 = 𝜀(𝑖)𝐷(1 − 𝜏) (5)
In this model, there is no paper currency. Money comprises only deposits (D). Equilibrium
in the money market is represented by a conventional LM curve in the y x i space. Money
supply (D, the model equivalent of M1) is given by the amount of reserves (R, the model
equivalent to M0), under the control of the central bank, multiplied by the money multiplier
(m):
𝑚(𝑖) = 1/[𝜀(𝑖)(1 − 𝜏) + 𝜏] (6)
Money demand (D) is quite conventional, depending on the interest rate and income (total
wealth is assumed constant and ipso facto eliminated). Therefore, equilibrium in the money
market is represented by the following LM curve, sloping upwards in the y x i plane:
𝐷(𝑖, 𝑦) = 𝑚(𝑖)𝑅 (7)
Having determined the money market equilibrium, we turn to equilibrium determination
in the remaining markets: loans, bonds and goods. The equilibrium in the loans market is given
by equation (8):
𝐿(𝜌, 𝑖, 𝑦) = 𝜆(𝜌, 𝑖)𝐷(1 − 𝜏) (8)
Given loan demand, L(ρ,i,y), and money demand, D(i,y), the nonbank public’s demand for
bonds is implicitly defined because total financial wealth is supposed constant. Finally, let’s
turn to the goods market equilibrium. It is summarized by an IS curve where the loan rate, ρ,
also enters:
𝑦 = 𝑌(𝑖, 𝜌) (9)
The key novelty of the BB model is precisely that ρ affects the IS curve. Since, by the
equilibrium in the loan market (equation (8)), ρ depends on D, which, in turn, by the
equilibrium in the money market (equation (7)), depends on R, monetary policy, i.e. the
amount of bank reserves (R), will also directly influence the goods market equilibrium .
The graphical representation is undertaken in the same familiar y x i plane, although a
tridimensional y x i x representation would probably be more instructive. To represent the
model in the y x i plane, we start by replacing D in the loans market equilibrium (equation (8))
by money supply, m(i)R, yielding:
𝐿 = (𝜌, 𝑖, 𝑦) = 𝜆(𝜌, 𝑖)𝑚(𝑖)𝑅(1 − 𝜏)
Then, the resulting equation can be solved to yield ρ as a function of the other variables: i,
y, R and τ:
𝜌 = Ø(𝑖, 𝑦, 𝑅, 𝜏) (10)
In (10), the derivative of ρ with respect to i is usually positive, because when i increases,
banks tend to allocate more of their free deposits to bonds, thereby lowering the amount of
8
loans. Given a downward sloped demand curve for loans, this will increase ρ. This effect is akin
to the substitution effect in consumer theory.
However, there is another effect, akin to the income effect. When i increases, the money
multiplier also increases, yielding more deposits from the same amount of bank reserves, R. If
this “income” effect is very strong, it may overcome the former “substitution” effect, and
make ρ a negative function of i.
Substituting the ρ, given by (10), into the goods market equilibrium condition (9), we get
the new IS, which is baptized in BB as the CC (“commodities and credit”) curve, in honor of the
late Don Patinkin.
𝑦 = 𝑌{𝑖, [Ø(𝑖, 𝑦, 𝑅, 𝜏)]} (11)
The CC curve is also downward sloping in the y x i plane, for the same reasons as the
typical IS curve. However, it now responds to shifts in R, as well as to shocks in the loan
market, affecting either the supply or the demand side. In the next section, this model will be
adapted so that it can account for sterilized interventions. It will then be used to evaluate the
effects of sterilized interventions under inflation targeting.
9
3. Effects of Sterilized Interventions
The BB model represents a closed economy, while the subject of this paper, sterilized
interventions, naturally suggests an open economy model. However, the introduction of a full
blown open economy model would distract us from the main goal of the paper: to show that
sterilized interventions (FX purchases) are generally expansionary even when they do not
affect (depreciate) the exchange rate. As already mentioned (see references in footnote 4),
both the theoretical and empirical evidences regarding the effectiveness of sterilized
interventions in affecting the nominal exchange rate are dubious. Of course, if sterilized FX
purchases depreciate the domestic currency, they would be, in most models, expansionary.
But what we aim to show here is another effect of sterilized interventions; even if FX sterilized
purchases do not depreciate the nominal exchange rate, they tend to be expansionary. The
expansionary effect studied in this paper is additional to the one derived from the possible
depreciation caused by sterilized FX purchases.
With that caveat in mind, we will proceed with minor adaptations of the BB model,
without explicitly introducing a foreign country or an exchange rate. The implicit assumption
will be that sterilized interventions will not affect the level of the floating exchange rate. As
previously noted, if they do, the expansionary effect of FX sterilized purchases would be even
stronger.
Sterilized interventions are usually defined as purchases or sales of FX that do not affect
the monetary base (R). However, in the inflation targeting framework, or any other monetary
policy framework in which the instrument is the interest rate (i) instead of a monetary
aggregate, the term sterilized interventions usually refers to FX transactions that do not alter
the interest rate that prevailed before the sterilized intervention (i*).10
Often, it is implicitly assumed that both definitions are equivalent, but it will be shown that
this is not the general case in models with a richer asset choice than the usual IS-LM one
between bonds and money.
Let us examine the mechanics of a sterilized FX purchase. For that, banks will be allowed to
have an alternative source of funding, foreign loans (FL), already denominated in domestic
currency units. For simplicity, these bank liabilities will not be subject to reserve requirements.
In order to account for foreign loans, equations (2), (3), (4), (5), (8), (10) and (11) of the
original model have to be modified in the following way.
𝐵𝑏 + 𝐿𝑠 + 𝐸 = 𝐷(1 − 𝜏) + 𝐹𝐿 (2’)
𝐿𝑠 = 𝜆(𝜌, 𝑖)[𝐷(1 − 𝜏) + 𝐹𝐿] (3’)
𝐵𝑏 = 𝛽(𝜌, 𝑖)[𝐷(1 − 𝜏) + 𝐹𝐿] (4’)
10 “Most central banks no longer target monetary aggregates, so instead, sterilized intervention can be thought of as foreign exchange market activity by the central bank that does not change its target interest rate” (Engel, 2013).
10
𝐸 = 𝜀(𝑖)[𝐷(1 − 𝜏) + 𝐹𝐿] (5’)
𝐿(𝜌, 𝑖, 𝑦) = 𝜆(𝜌, 𝑖)[𝐷(1 − 𝜏) + 𝐹𝐿] (8’)
𝐿(𝜌, 𝑖, 𝑦) = 𝜆(𝜌, 𝑖)[𝑚(𝑖)𝑅(1 − 𝜏) + 𝐹𝐿] (8a’)
𝜌 = Ø(𝑖, 𝑦, 𝑅, 𝜏, 𝐹𝐿) (10’)
𝑦 = 𝑌{𝑖, [Ø(𝑖, 𝑦, 𝑅, 𝜏, 𝐹𝐿)]} (11’)
The interpretation of the expansionary effect of a sterilized FX purchase is the following.
Assume that a FX inflow enters the economy as foreign loans to banks. FL is the equivalent
amount of the foreign loans in domestic currency at the prevailing exchange rate, assumed to
be unaltered by the sterilized purchase. As previously explained, if sterilized purchases were
effective in depreciating the exchange rate, their expansionary effect would be even stronger.
First, the CB purchases all the foreign currency and issues domestic currency (R). Second,
the CB soaks up the newly issued domestic currency, exchanging it for government bonds. The
resulting bank sector balance sheet is shown in Figure 2.
Figure 2: Representative Bank Balance Sheet After the Sterilized FX Purchase
Bank Balance Sheet Assets Liabilities
R (bank reserves) D (deposits)
Bb + FL (bonds) FL (foreign loans)
Ls (loan supply)
Remember that the sterilized FX purchase under inflation targeting is supposed to restore
the interest rate to its previous level. Therefore, the asset allocation in Figure 2 cannot
represent an equilibrium for the bank with the same rates i and that prevailed before the
sterilized FX purchase. To view this, compare Figure 2 with Figure 1. Figure 2 shows that the
new bank liabilities, FL, were fully allocated to bonds. None was allocated to loans. For this to
be an equilibrium for the bank, at the previous interest rate, i, the loan interest rate, , must
have fallen. And, with a fall in the loan interest rate, , loan demand must have expanded. In
equilibrium, loan supply would also expand, provoking an expansion in output.
In fact, the sequence of events is the following. In the first stage of the sterilized FX
purchase, the CB purchases the foreign exchange and delivers the equivalent amount in
domestic currency (at the prevailing exchange rate) to the bank. This money injection causes
both i and to fall, shifting both the LM and the CC curves to the right, with [E’] being the new
equilibrium, as shown in Chart 1. Given this displacement of the CC curve, the resulting
interest rate is always higher than would be the case in the traditional IS-LM model ([E’’]),
where the IS curve does not respond to changes in R.
11
Chart 1
As defined in most textbooks, a sterilized intervention would be completed by a
contractionary open market operation that would fully offset the increase in R, bringing back
the equilibrium to [E]. However, in the inflation targeting framework, what the CB has to do is
to restore i to its previous level. Given the change in the CC curve, this is obtained with a
smaller sale of bonds than would be the case in the IS-LM model.11 Chart 2 displays the
equilibrium ([E’’’]) at the end of the sterilized FX purchase that restores the previous interest
rate (i*). Note that LM2 remains to the right of the original LM0, showing that not all money
issues were removed by the sterilization procedure that restored the original interest rate.
The size of the contractionary open market operation needed to shift the interest rate (i)
back to the level determined by the MPC is always smaller than the amount necessary to bring
R back to its previous level. One way to understand why is to note that, in the second stage of
the sterilized FX purchase, the sterilization itself, what the CB does is to replace reserves by
bonds in banks’ assets. This operation contracts both the LM and the CC curves. However,
given their larger liabilities, the banks, facing the same i as before the sterilized FX purchase,
now provide more (and cheaper) loans, thereby expanding output. The final asset allocation
for the bank will have higher loan supply, as well as higher bank reserves.
Another way to appreciate how the interest rate rule leads to this incomplete sterilization
is the following. Imagine that, after the sterilized FX purchase is completed, the bank sector
balance sheet looked like the one in Figure 2. As argued before, this cannot qualify as
11 In fact, BB calls attention to the possibility that “… a rise in bank reserves might conceivably raise the
rate of interest in the credit model” (Bernanke and Blinder (1988), p. 437). If this were the case, the central bank would have to conduct an expansionary open market operation to shift the interest rate (i) back to the level determined by the MPC.
- +
12
equilibrium with the same rates i and that prevailed before the sterilized FX purchase.
Therefore, banks will sell bonds to generate funds to make more loans. This portfolio
adjustment tends to increase the interest rate and decrease the loan rate. As the inflation-
targeting CB counteracts the interest rate increase with expansionary open market operations,
the amount of bank reserves is increased.
This latter interpretation has the advantage of better conveying the timing of the
expansionary effect in this static model. After all, sterilized interventions are financial
procedures that are conducted in a matter of minutes. Therefore, it is not reasonable to
assume that output would expand and contract in such a short period of time. However, this is
not the correct interpretation of the timing behind this static model. As explained in the
previous paragraph, once the sterilized FX purchase has been completed, and i is restored to
its previous level, the bank is not in equilibrium (see Figure 2), and will substitute loans for
bonds in its portfolio. As it does that, it pushes i up and down. The inflation-targeting CB
purchases the bonds and issues money to keep i at its target level, thereby monetizing the
economy. All these events, that may take days or weeks, eventually bring the economy to its
new equilibrium ([E’’’] in Chart 2). In summary, the timing of the effect, displayed in Chart 2,
has to do with the speed with which banks reallocate their portfolios from bonds to loans, not
with the speed of the sterilized intervention itself.
Bank reserves increase because, as output increases, the previous rate of interest is
restored at a higher level of money demand, which, in equilibrium, equals money supply. In
other words, the higher money supply is needed, in equilibrium, because, with higher y and
the same i, money demand increased after the sterilized FX purchase. That is, with a higher y,
the CB does not have to mop up all the money it had previously issued to restore the interest
rated, i. After the sterilized intervention, i is back to its previous level, but y is larger. This
occurs because there is more and cheaper credit in the economy. Given the shift in CC, due to
more and cheaper credit, to restore the initial level of output, y*, the CB would have to raise i
above the initial level i*.
Even if the CB were to remove all the money used to purchase the FX inflow, a small
expansionary effect would still occur. This may be seen in Chart 3. If the sterilization were to
remove all money issued, the previous LM curve would be restored. This is why LM3 lies on top
of LM0. However, given increased bank liabilities, the portfolio effect makes the CC3 curve
remain to the right of the original one, CC0, before the FX inflow, with more and cheaper loans
being provided. Therefore, output still expands when compared to its original level (compare
the new equilibrium, EIV, with the original one, E).
13
Chart 2
Chart 3
For the sake of thoroughness, it is useful to examine one last case, when the CB contracts
monetary policy in order to eliminate any expansionary effect on output. As shown in Chart 3,
because larger bank liabilities increase the amount of loans and decrease the loan rate, the CC
- +
- +
14
remains to the right of the original one when the original LM is restored. Therefore, in order to
restore the original output level, it is necessary to over-sterilize, as displayed in Chart 4. Note
that LM4 lies to the left of the original LM0. The new equilibrium, EV, has the original output
restored. The interest rate is higher than the original one. It is also higher than the one that
would have prevailed had the CB fully sterilized, in the textbook sense of mopping up all
money issued.
In sum, with monetary policy being conducted via an interest rate rule, as is the case in the
inflation targeting framework, sterilized FX purchases are expansionary. Even if the CB were to
fully offset the money issuance, a residual expansionary effect would still occur, because of the
portfolio effect on banks’ balance sheets.
Of course, whether or not such effects are of practical importance is an empirical issue. In
Section 5, empirical evidence will be provided in order to argue that this mechanism may have
played an important role in propping up aggregate demand in Brazil. However, before we
examine the empirical evidence, Section 4 develops the model in order to derive another
important policy conclusion regarding policy measures to counteract the detrimental effects of
capital inflows.
Chart 4
- +
15
4. Different Types of Capital Inflows and Policy Measures
In the discussion regarding how to react to excessive capital inflows, it has been argued
that receiving countries would do well to reduce interest rates, in order to attract less so-called
hot money. In fact, the Central Bank of Turkey, on December 16, 2010, cut interest rates amid
rising inflation and a low output gap. The deputy governor of the Central Bank of Turkey,
Erdem Basci, argued that gradual rate cuts were the best way to prevent excessive capital
inflows fuelling asset bubbles and currency appreciation.12
The model described in the previous section may be used to analyze whether or not such
policy prescriptions are warranted. In order to do so, it is useful to differentiate two kinds of
capital inflows: those fully destined to the direct purchase of domestic government bonds and
all the others. The latter category includes all kind of inflows that, one way or another, will
fund increases in aggregate demand. Those flows will be absorbed by private firms and
financial institutions or even public institutions (including government-owned banks). Those
are the flows that were analyzed in Section 3’s model. The inflows that enter the country to
directly purchase government bonds do not have the expansionary effect described in this
paper, since they neither generate an expansion in base money, nor the portfolio effect
necessary to move the CC curve. A good example would be a special fund (SIV) set up
exclusively to buy government bonds for foreign investors. These inflows represent an external
source of demand for government bonds, thereby creating a downward pressure on interest
rates (therefore increasing government bond prices). This could, indirectly, increase aggregate
demand if the CB did not act to keep interest rates constant, but it does not create the
expansionary effect through the expansion of bank liabilities.
Both kinds of inflows are associated with currency appreciation, but only the latter form
cause credit to expand. And the capital inflows that fund bank credit are attracted by the high
, not the high i. Therefore, advocating reductions in interest rates to deal with an excessive
capital inflow situation may help to deter only the capital flows that are attracted by the high
interest rate differential of government bonds.
However, the lower interest rate will probably increase the expansionary effects of the
inflows that fund bank credit, because the lower interest rate will fuel aggregate demand,
through the usual interest rate channel. Remember that, in this model, the decline in the
interest rate is brought about by an increase in base money, R, which expands both the LM
and the CC curves, leading to a larger expansion in output, y. The fall in i tends to reduce , but
the expansion on y, by increasing the demand for loans, mitigate this effect. Even if the
elasticities are such that the fall in i substantially reduces , thereby mitigating capital inflows,
the final result may be worse, in terms of increasing aggregate demand, than what would have
happened in the absence of the monetary loosening . As shown in Chart 4, to eliminate the
expansionary effect of sterilized FX purchases, in the absence of a fiscal contraction, it is
necessary to increase interest rates, not reduce them.
Even if so-called macroprudential measures (increases in , in the model) are deemed
adequate for deterring credit expansion, the fall in interest rates will increase macroeconomic
12 Financial Times (2010).
16
policy’s dependence on them. The high current inflation in Turkey (10.45% in January, 2012)
shows that the strategy of lowering the interest rates in face of massive capital inflows is not a
sensible one.
17
5. FX Sterilized Interventions in Brazil and Money
After the peak of the 2008 crisis, Brazil resumed sterilized FX purchases as early as
February, 2009. Since then, foreign reserves have risen from USD 187 bi to USD 350 bi,
surpassing 15% of GDP in October, 2011.
Chart 5 shows that the monetary base has also expanded rapidly. In 2010, it increased
25%, or BRL 40 bi, compared to an inflation rate of 6%. Real GDP expanded by 7.5%.
As shown in Chart 5, FX purchases (almost BRL 80 bi in 2010) were one of the main factors
accounting for such a robust increase in money. Of course other CB operations affected the
monetary base, and it is very hard to show causality, but, prima facie, it seems plausible that
the story told by the previous model explains at least part of what has been happening in
Brazil.
Chart 5
Source: Central Bank of Brazil
Another important piece of evidence that suggests that the story behind this model might
be relevant is the behavior of credit markets in Brazil. Chart 6 makes it clear that, albeit very
expensive, bank credit has been expanding in Brazil while the average credit interest rate has
been declining. This is compatible with a supply of credit expansion larger than the increase in
credit demand, precisely what the model presented here predicts would happen with massive
sterilized interventions.
127
140
153
166
179
192
205
218
-40
-20
0
20
40
60
80
100
120
140
Monetary Base and Its Main Expansion Factors(12-month-moving average; BRL bi)
Monetary Base - M0 (RHS) FX purchases (LHS)
MacroprudentialMeasures
18
Chart 6
Source: Central Bank of Brazil
Furthermore, Chart 7 shows that the rate on loans to individuals follows the one-year-
interbank rate almost perfectly, with a three-month lag.13 This is quite reasonable, since the
interbank rate is the best proxy for banks’ cost of funding. However, since the beginning of
2010, this positive correlation seems to have broken down: while the interbank rate rose, the
loan rate kept following it until November,14 the month prior to the imposition of
macroprudential measures to deter credit growth.15 The interbank rate follows expectations
regarding the interest rate set by the Central Bank, the Selic, equivalent to i in the model. The
loan rate is the equivalent of the in the model. The model asserts that, under massive
sterilized FX purchases, the highly unusual negative correlation, as observed in 2010, should be
the outcome.
Econometrically, the best result is obtained when the loan rate to individuals is regressed
against its marginal cost, the one-year-interbank rate, lagged three-months, together with FX
purchases by the CB (12-month average) and a dummy for 2010 interacted with the FX
purchases. Table 1 displays the results. FX purchases by the CB become statistically significant
only when the 2010 dummy is included in the regression (interacted or not with FX purchases).
These econometric results are compatible with the main lessons from the model: the
resumption of FX purchases, after recovering from the 2008 crisis, kept the loan rate falling
13 In Chart 4, the interbank rate is leaded three months. 14 I wish to thank Eduardo Loyo for pointing this out to me. 15 The effects of macroprudential measures may also be observed at the end of Chart 6, when credit volume stops growing and the average credit interest rate increases.
700.000
740.000
780.000
820.000
860.000
900.000
940.000
980.000
1.020.000
1.060.000
30
32
34
36
38
40
42
44
% a.a.
Total Credit Freely AllocatedQuantity and Interest Rate
Total Credit (RHS) Average Interest Rate (LHS)
Macroprudential Measures
19
even when the interest rate was rising. That lasted until macroprudential measures were
implemented in December, 2010. The Brazilian economy therefore performed as predicted by
the model, with the increase in sterilized FX purchases in 2010 causing the loan rate to fall
despite the increase in its marginal cost - the one-year-interbank interest rate.
Chart 7
Source: Central Bank of Brazil
In order to duly account for non-stationarity, endogeneity and autocorrelated errors,
the econometric analysis was performed using the Phillips-Hansen (1990) procedure. Table 2 displays the results. The FX purchases make the loan rate fall relatively to its long term equilibrium with the interest rate. That effect is even more pronounced in 2010. Note how the coefficients are very similar to the ones in the OLS regression, denoting that the results are quite robust.
34
42
50
58
66
74
82
90
7
11
15
19
23
27
31
% p
er y
ear
% p
er y
ear
Interbank Market Rates and Loan Rates
1-year interbank market rates (t-3) Rates on loans to individuals (t)
Macroprudential Measures
20
TABLE 1: The Loan Rate and Sterilized Interventions
Dependent Variable: Loan rate to individuals
Independent Variables: One-Year-Interbank rate, FX Purchases (12-month average) and FX Purchases (12-month average) multiplied by a Dummy for 2010
Sample: 2000:12-2011:08 (T = 129)
Variable Coefficient Stand. Error t-Ratio P-Value
Constant 30.0487 1.28187 23.44 1.39e-047 ***
One-Year-Interbank Rate (t-3) 1,85388 0,0655097 28,30 3.40e-056 ***
FX Purchases (12-month average)
-0,0392121 0,00620752 -6.317 4.27e-09 ***
FX Purchases (12-month average) multiplied by a Dummy for 2010
-0,0877712 0,0122744 -7,151 6,38e-011 ***
Mean dependent variable 58,14132 S.D. dependent var. 12,53158 Sum squared resid. 1055,908 S.E. of regression 2,906418 R-squared 0,947470 Adjusted R-squared 0,946210 F(3, 125) 751,5366 P-value(F) 9,26e-80 Log-likelihood -318,6443 Akaike Criterion 645,2885 Schwarz Criterion 656,7278 Hannan-Quinn 649,9365
0,499275 Durbin-Watson 0,975744
TABLE 2: Cointegration Regression
Dependent Variable: Loan rate to individuals
Independent Variables: One-Year-Interbank rate, FX Purchases (12-month average) and FX Purchases (12-month average) multiplied by a Dummy for 2010
Sample (adjusted): 14 144 Included observations: 131 after adjustments Cointegrating equation deterministics: C Long-run covariance estimate (Prewhitening with lags = 1 from HQ maxlags = 5, Quadratic-Spectral kernel, Andrews bandwidth = 1.1543)
Variable Coefficient Stand. Error t-Ratio Prob
Constant 28.26229 2.396402 11.79364 0.0000
One-Year-Interbank Rate (t-3) 1.977465 0.122784 16.10528 0.0000
FX Purchases (12-month average)
-0.037931 0.011613 -3.266133 0.0014
FX Purchases (12-month average) multiplied by a Dummy for 2010
-0.105830 0.024147 -4.382788 0.0000
Mean dependent variable 58.31939 S.D. dependent var. 12.51822 Sum squared resid. 1261.924 S.E. of regression 3.152207 R-squared 0.938055 Adjusted R-squared 0.936592 Long-run variance 29.80132 Durbin-Watson 0.851448
21
The model also predicts that deposits should increase in times of large sterilized FX purchases.
Note that Brazilian law forbids deposits in foreign currency in Brazilian financial institutions.
Chart 8 shows the increase of demand deposits and total deposits in Brazilian banks. Chart 9
displays the same data in % of GDP. It shows that they increased sharply as a % of GDP, until
the crisis, and then stabilized, as a % of GDP. Furthermore, Granger-causation runs from the
sterilized FX purchases to loans, but not the other way around.
However, when we look at all the liabilities of the banking sector (everything, except
capital), we see much higher growth since 2006, as a % of GDP (see Chart 10). Therefore, there
is evidence that banks liabilities have been growing, as predicted by the model. However, this
phenomenon has been happening since 2006, and was, apparently, not affected by the 2008
crisis.
Chart 8
Source: Central Bank of Brazil
0
50
100
150
200
250
0
200
400
600
800
1.000
1.200
1.400
1.600
1.800
Bill
ion
R$
Bill
ion
R$
Total Deposits and Demand Deposits
Total deposits Demand deposits
Macroprudential Measures
22
Chart 9
Source: Central Bank of Brazil
Chart 10
Source: Central Bank of Brazil
3%
4%
5%
6%
7%
25%
30%
35%
40%
45%
Deposits / GDP
Total Deposits Demand Deposits
Macroprudential Measures
Macroprudential Measures
60%
70%
80%
90%
100%
110%
120%
Bank Liabilities / GDP
MacroprudentialMeasures
23
Chart 11
Source: Central Bank of Brazil
As mentioned before, several policy measures, besides sterilized interventions, were taken
to deal with what were perceived to be excessive capital inflows. Another set of policy
measures have targeted a subset of capital inflows. Since March 2011, short-term credit (up to
two years) obtained by Brazilian banks abroad has been taxed to deter further capital inflows.
These measures aim to deter the FX flows that prompt sterilized interventions.
An additional policy action was the decision taken in 2010 to let the Brazilian Central Bank
return to derivatives markets, trading with currency swaps. The purchase of a currency swap
by the Central Bank is equivalent to a sterilized intervention,16 in the sense of interventions
that keep money constant. As the previous results show, what are generally referred to as
sterilized interventions under inflation targeting are not equivalent to the textbook definition.
Therefore, the two forms of intervention may produce different results in terms of their
effectiveness in altering the exchange rate.
Since 2010, something seems to have changed in the behavior of the BRL/USD exchange
rate. Although there has been a historical correlation between terms of trade and the
exchange rate, improvements in the terms of trade since mid-2010 have not been
accompanied by a corresponding appreciation of the BRL, thereby transmitting the
commodities’ positive price (in USD) shock to domestic inflation. Chart 12 shows this
phenomenon.
16 Obstfeld and Rogoff (1996), pp. 597-9.
50%
60%
70%
80%
90%
100%
110%
Total Assets and Liabilities (BNDES not included)
Assets - Interbank Loans (%GDP) Liabilities (%GDP) Liabilities - Interbank Loans (%GDP)
MacroprudentialMeasures
24
Chart 12
In addition, the volatility of the BRL/USD rate has fallen markedly, as shown in Chart 13.
These are evidences that government interventions may be affecting the exchange rate,
mitigating the appreciation that would occur under free floating.
Chart 13
0,20
0,28
0,36
0,44
0,52
0,60
0,68
0,76
0,84
60,00
75,00
90,00
105,00
120,00
135,00
150,00
165,00
180,00
BR
L/U
SD
Ind
ex
Exchange Rate and Commodities Prices
Exchange Rate CRB Index (in USD) CRB Index (in BRL)
0
0,01
0,02
0,03
0,04
0,05
0,06
-0,1
-0,05
0
0,05
0,1
0,15
Vo
lati
lity
Dai
ly C
han
ges
Exchange Rate (BRL/USD): Daily Changes and Volatility
Daily Changes in the BRL/USD Exchange Rate (LHS) Exchange Rate Volatility (20-day Rolling Window)
25
As previously noted, the theoretical and econometric evidences regarding the
effectiveness of sterilized interventions in altering the exchange rate are mixed. However, this
paper argues that the kind of sterilized intervention conducted in Brazil does not correspond
to the canonic definition of sterilized intervention. The sterilization of the increase in money
caused by FX purchases is only partial, because the expansionary effect on output increases
money demand, thereby requiring less monetary contraction to return the interest rate to its
previous level. In a nutshell, Brazil has not been fully sterilizing its FX purchases, in the
“monetarist” sense of the expression. Therefore, it does not come as a surprise that FX
purchases may have been more efficient in mitigating nominal exchange rate appreciation.
In sum, this section has presented empirical evidence showing that many phenomena
observed in the Brazilian economy are compatible with the model’s predictions. Massive
sterilized interventions have not been neutral; they increased banks´ liabilities and the supply
of credit, making credit cheaper and more abundant, even when the Central Bank of Brazil was
raising interest rates. This, in turn, expanded aggregate demand, making it harder to keep
inflation at bay. The next section will summarize this paper’s main conclusions and policy
prescriptions.
26
6. Concluding Remarks
Many countries have resorted to sterilized FX purchases to mitigate exchange rate
appreciation and large credit expansions caused by massive capital inflows. Sterilized FX
interventions are defined as FX purchases (sales) by the CB followed by open market
operations that offset their monetary impact.
Under inflation targeting, or any monetary policy regime with an interest rate rule,
sterilized FX interventions usually refer to FX operations followed by open market operations
that restore the interest rate to its target. Restoring the interest rate to its previous level may
not be equivalent to restoring the monetary base to its previous level.
We adapt a simple model17 with a banking sector and richer asset structure than just the
money and bonds present in the classical IS-LM model in order to argue that, in general, FX
sterilized interventions under inflation targeting are expansionary.
When bank credit is explicitly introduced into the IS-LM model, increases in the monetary
base (bank reserves) affect not only the LM curve, but also the new IS curve, termed CC, for
“commodities and credit”. This effect is caused by bank loans, which become cheaper and
more abundant when bank deposits rise because of the increase in bank reserves. Therefore,
when credit is incorporated into the model, monetary policy, by affecting banks’ behavior,
becomes more powerful. An increase in bank reserves will lead to a larger output expansion
than in the usual IS-LM model.
We use this model to argue that sterilized interventions under inflation targeting is
expansionary. When a foreign loan is taken out by a bank, its liabilities increase. The sterilized
FX purchase by the CB is aimed at making the bank hold all the increase in liabilities in the form
of government bonds. However, with increased liabilities, the bank wants to diversify its
holdings, and channels part of the new funds into loans. This pressure to reallocate the bank´s
portfolio—when the assets (bonds and loans) are imperfect substitutes, and given that the
previous interest rate has been restored after the sterilized FX purchase—increases loan
supply, lowers the loan rate, thereby expanding aggregate demand.
The model in this paper relies on a portfolio balance effect generated inside the bank.
Recent research on the behavior of financial institutions has shown that they tend to over
leverage in good times.18 The external funding provided by capital inflows constitutes one
important way through which this leverage may occur. As explained by the model, such an
effect occurs despite sterilization.
On the other hand, the expansionary effect of sterilized interventions has other
transmission channels apart from the one described in the model, which, in this case, is
constituted by the banking sector. The main idea is that capital flows will increase aggregate
demand when the CB keeps the interest rate constant at its level before the capital flows and
the sterilized FX purchases. This is true for banks that fund their domestic loans by borrowing
17 Bernanke and Blinder (1988). 18 Adrian and Shin (2009).
27
from abroad, but is also valid for FDI or corporate securities issued abroad to fund investment
projects.
It is also true for trade flows that allow firms that export to fund their investment projects.
For example, if an exporter decides to undertake an investment project, and funds it with its
export proceeds, because, for example, the returns are higher than those obtained by
investing these export revenues in governments bonds, at prevailing interest rates (that will be
kept constant with the sterilized intervention), there will be an expansionary effect despite the
sterilization. This expansionary effect is akin to the effect displayed in the model in Section 3.
However, the expansionary effect does not hold for capital flows directly targeted to
purchase government bonds (e.g., carry-trade), for they would not increase aggregate
demand. Capital flows directly targeted to purchase government bonds represent an external
source of demand for government bonds, thereby creating a downward pressure on interest
rates (therefore increasing government bond prices). If the CB did not act to keep interest
rates constant, these capital inflows could lead to an increase in aggregate demand, but it
could not create the expansionary effect through the expansion of bank liabilities.
The part of the intuition that explains why money expands when the interest rate is kept
constant is the following. Capital flows not directly targeted to government bonds raise
aggregate demand, thereby also increasing money demand, at the prevailing interest rate.
Therefore, money supply has to increase, in equilibrium. In the model, an increase in the
monetary base is what leads, in the first place, to an increase in loans. Therefore, sterilized
interventions under inflation targeting—not fully sterilized in the sense of keeping money
constant—become expansionary.19 At the new equilibrium there will be higher aggregate
demand, a higher quantity of money, lower loan rate and higher quantity of loans at the same
interest rate. The timing to arrive at the new equilibrium has to do with how fast banks
reallocate their portfolios after the sterilized intervention, not with the few minutes the CB
takes to perform a sterilized FX purchase.
Brazil’s recent experience was reviewed to argue that the expansionary effect of sterilized
interventions may be significant. The monetary base expanded 25% in 2010, while GDP grew
7.5%. Credit also increased substantially, with most loans becoming cheaper. Besides timid
increases in interest rates, at the end of 2010, the government has resorted to
macroprudential measures, such as increases in reserve requirements. All these evidences are
compatible with the expansionary effect of sterilized interventions under inflation targeting.
One empirical evidence that the mechanism behind the model may be important to
explain what happened in Brazil is the joint behavior of the consumer loan rate and the
banks´marginal cost for loans, i.e., the one-year-interbank interest rate. They showed a
remarkably high correlation, as expected, until the end of 2009, a time when both rates were
falling. Since then, the interest rate has gone up, anticipating the increases signaled by the
Central Bank of Brazil, that were later actually implemented. However, the loan rate kept
19 The model shows that even if the CB would were to fully sterilize, a smaller expansionary effect would occur. This is because, with larger liabilities, banks will offer more and cheaper credit, thereby expanding the CC curve.
28
falling. This unusual negative correlation is precisely what the model says would happen under
massive sterilized FX purchases. Econometric investigation confirms the intuition.
The main policy implication of this paper is that when a country receives large inflows of
FX that are not aimed at purchasing government bonds (including trade revenues), it is not
sufficient, in order to fully sterilize FX purchases, for the CB to restore the previous interest
rate level. If the FX flows affect aggregate demand, e.g. via bank credit, sterilized interventions
under inflation targeting will be expansionary.
Another policy implication is that policy strategies like the one adopted in late 2010 by
Turkey, that combine lower interest rates with the so-called macroprudential measures, are
inconsistent. The capital inflows that would be deterred by the fall in interest rates, those
aimed at purchasing government bonds, are not the ones that make FX sterilized purchases
expansionary. Therefore, the dependence on the so-called macroprudential measures to keep
inflation at bay would be even higher. The increase in inflation in Turkey corroborates the idea.
Only a model that differentiates the loan rate from the interest rate on government bonds, as
the one in this paper, can be useful to derive policy implications for these strategies, very
much in fashion among emerging markets, until recently, when capital flows turned back.
When and if capital resumes its flow toward emerging markets, inflation targeting
countries that conduct sterilized interventions to mitigate the appreciation of the exchange
rate in face of massive capital inflows will have another reason for concern. Even if those
sterilized interventions are effective in preventing nominal exchange rate appreciation, they
may represent a positive shock to aggregate demand, thereby increasing inflation with all its
detrimental effects, among them, the appreciation of the real exchange rate.
29
7. References
Adrian, T. and H. Shin, 2009. The Shadow Banking System: Implications for Financial
Regulation, Federal Reserve Bank of New York Staff Reports, nº 382, July.
Bernanke, B. and A. Blinder, 1988. Credit, Money, and Aggregate Demand, American Economic
Review, May.
Central Bank of Brazil, 2011. Inflation Report, page 14. June. Available at
http://www.bcb.gov.br/htms/relinf/ing/2011/06/ri201106c6i.pdf.
Central Bank of Chile, 2011. Financial Stability Report, page 18. First Half 2011. Available at
http://www.bcentral.cl/eng/publications/policies/pdf/fsr1_2011.pdf.
Central Bank of Turkey, 2011. Inflation Report 2011-III, pages 11/12. Available at
http://www.tcmb.gov.tr/yeni/eng/.
Credit Suisse, 2011. Brazil Economics Digest, October 20. Available at http://www.credit-
suisse.com/researchandanalytics .
Dominguez, K. and J. Frankel, 1993. Does Foreign Exchange Intervention Work?. Washington,
DC: Institute of International Economics.
Engel, Charles, 2013, “Exchange Rates and Interest Parity”, NBER Discussion Paper #19336.
Financial Times, 2010. Turkey surprises with interest rate cut, by Delphine Strauss, December
16.
Mishkin, F., 2000. Inflation Targeting in Emerging Market Countries. American Economic Review,
vol. 90(2), pages 105-109.
Neely, C., 2005. An Analysis of Recent Studies of the Effect of Foreign Exchange Intervention. Federal
Reserve Bank of Saint Louis Working Paper No. 2005-030B.
Phillips, P. and B. Hansen, 1990. Statistical Inference in Instrumental Variables Regression with I(1)
Processes. Review of Economic Studies, vol. 57, pages 99-125.
Obstfeld, M. and K. Rogoff, 1996. Foundations of International Economics, MIT Press,
Cambridge, MA, US.
Sarno, L. and M. Taylor, 2001. Official Intervention in the Foreign Exchange Market: Is it
Effective? If so, how does it work, Journal of Economic Literature, 39(3): 839-868.